Introduction
Chronic obstructive pulmonary disease (COPD) is characterised by poor and worsening lung function, and significant patient and societal burdens [
1]. Prevention and treatment of exacerbations have been identified by the Global Initiative for Chronic Obstructive Lung Disease (GOLD) as a priority since they are associated with lung function decline. The recommendation of inhaled corticosteroid/long-acting β
2 agonist combinations (ICS/LABA) is based on evidence of a reduction in exacerbations in studies such as TORCH [
2]. Re-analyses of studies such as the TORCH have cast doubt on whether the corticosteroid component of these combinations, is efficacious [
3]. However, ICS and ICS/LABA combinations are commonly prescribed inappropriately early in the course of the disease where bronchodilator therapies (short acting bronchodilators, long-acting muscarinic antagonists (LAMAs), LABAs) may have been more suitable [
4,
5]. The recent results of the FLAME study call this strategy into further question [
6]. Not only has the efficacy of ICS in COPD, especially mild-to-moderate disease, remained open to question [
4], but their use especially at the high doses recommended in COPD are now associated with significant adverse events such as pneumonia, cataracts, glaucoma, accelerated bone turnover and diabetes [
7]. These are acknowledged in current management strategies and in guidelines to a varying extent, although there is no clear guidance on how the increased risk should inform decision-making by physicians [
1,
8].
More recently, with respect to pneumonia, there has been debate as to whether all ICS carry a similar risk profile and whether these are dose-related. The majority of studies demonstrating increased pneumonia risk were performed with fluticasone propionate (FP) doses (500–1000 µg/day), including the 3-year TORCH and 2-year INSPIRE studies [
2,
9‐
13]. Meta-analyses of randomised studies have reported an increased pneumonia risk of up to 70% with ICS use [
14,
15]. However no or lower risk of pneumonia have been noted with budesonide compared to FP [
16,
17]. Similar observations have been made from retrospective database analyses, and observational matched primary care medical records review [
18,
19]. Of the studies of budesonide and beclomethasone in COPD, only two have demonstrated a significant increase in reported pneumonia [
20,
21].
Despite the growing evidence that FP-based therapies have higher rates of pneumonia criticisms have been raised including the lack of clear clinical information from databases, poor patient-control matching, the retrospective nature of all analyses, and the mix of protocol designs comparing ICS/LABA versus their monotherapies and/or placebo. Indeed, the view of the European Medicines Agency, which adopted the Committee for Medicinal Products for Human Use, was that there was no difference in interclass respiratory adverse events between ICS [
22]. In 2008, the then largest randomised controlled study of COPD patients, the UPLIFT study, was published [
23]. An unusual feature of this 4-year study was that patients were maintained on existing non-cholinergic inhaled therapies whilst being randomised to tiotropium or matched placebo. We hypothesised that a comparison of patients taking ICS at entry to the study would allow comparison of the rates of adverse respiratory events in patients taking FP compared to those on other ICS.
Discussion
Our analysis of this large 4-year prospective randomised study has established that ICS use at baseline is associated with an increase in both pneumonia (>20%) and COPD exacerbations (45%) compared to those not on ICS. The ITT principle was used in these analyses because of the difficulties in ascertaining the use and adherence to ICS therapy over a duration of the 4 years of study. ITT, although not ideal, was considered the safer analytic method in this secondary endpoint dataset. To support our assumptions, we conducted a number of sensitivity analyses without evidence of any interactions (see supplementary methods). As anticipated, the two groups were not matched at baseline in severity, with a quarter of ICS subjects in GOLD stage IV vs 19% in the no ICS group. Similarly FEV1% predicted in ICS group was 38% compared to 41% in the no ICS group. Thus the uneven distribution of baseline characteristics make it impossible to determine whether the differences in outcome are due to the adverse effects of treatment or the greater disease severity in those receiving the more intensive treatment.
In contrast, the 2 subgroups of patients receiving ICS that is, FP or other ICS, were very well matched in all the baseline characteristics. Thus, it is legitimate to compare the two treatment regimes on an intention-to-treat basis. In this study, we found a significant increase in the rate and number of cases of pneumonia and a significant increase in the number, but not the time-to-first event of COPD exacerbations. This suggests that any effect was disproportionately exhibited in those with frequent exacerbations. As stated earlier, UPLIFT differs from other therapeutic trials in COPD in that concomitant medications were permitted throughout [
23]. Thus, the increase of nearly a third in the episodes of pneumonia seen in our analysis poinst to an adverse effect of FP therapy rather than other factors such ICS withdrawal as seen in other studies. Indeed, when the frequency of pneumonia in the no ICS group is compared to the other ICS, rates are similar. Thus, almost all of the excess of pneumonia episodes due ICS treatment can be ascribed to FP (Table
2).
Episodes of COPD exacerbations were also increased in the FP group. However, the time-to-first exacerbation was initially similar across the 3 groups, before diverging with clear increase for both the ICS groups (Fig.
2). Given that there was less severe disease in the no ICS group, this suggests that time-to-first event may be a less good marker of outcome than frequency of exacerbation. We suggest that ICS, and particularly FP as stated above, may adversely affect those prone to recurrent exacerbations of COPD: Contrary to many guideline recommendations [
1,
25].
Our findings confirm the observations of several randomised trials of varying durations and doses of FP. However, unlike the 2-year INSPIRE and 3-year TORCH studies where high HRs of pneumonia were reported [
10,
11], our analysis demonstrates a significant but more modest effect. It is also much lower than other shorter duration randomised trials and of studies with lower FP doses [
12,
13]. We suggest that this may be due to the longer observation period of 4 years in UPLIFT. Retrospective database and case–control analyses of longer durations have also reported higher rates of pneumonia [
18,
19].
Our observations are similar to the retrospective case–control PATHOS study which found lower rates of COPD exacerbations and pneumonia in patients treated with budesonide/formoterol compared to FP/salmeterol [
19,
26]. In a study of dual bronchodilators (LAMA/LABA combination) versus FP/salmeterol, the latter also resulted in significant increases in pneumonia and COPD exacerbations in moderate-to-severe patients [
27]. The superiority of dual bronchodilation over ICS/LABA has also been confirmed in the recent FLAME study [
6].
Fluticasone differs from other ICS, such as beclomethasone and budesonide, in the presence of a fluorine moiety. This drags the electrons across the molecule altering not only potency but also lipid solubility [
28,
29]. In vivo this is characterised by an alteration in the volume of distribution, fluticasone dwelling for longer in the lipid membranes. The consequence of this is a slower clearance from the lungs and other tissues. This will clearly have effects on lung immunity and epithelial barrier function [
30]; thus, dampening down inflammatory responses. Alternatively, potent steroids may increase the risk of potential aspiration events. That fluticasone particularly prone to these adverse effects is confirmed by recent reports of increased pneumonia with fluticasone furoate in COPD [
31,
32]. Our analysis does not exonerate other ICS from adverse respiratory outcomes; it merely suggests that these are more prevalent with fluticasone. This is supported by a recent study suggesting that withdrawal of ICS decreases the risk of pneumonia and that this is particularly marked with FP [
33]. In ICS withdrawal study, WISDOM, a short-term increase in adverse respiratory events was observed on gradual steroid withdrawal from triple inhaled therapy [
34]. In contrast, in the FLAME study, abrupt withdrawal of ICS did not increase exacerbation rates [
6]. Asthma COPD overlap syndrome (ACOS) [
35] may explain some of the anomalies seen in these studies.
A clear observation from this analysis is that long-acting bronchodilator therapy in the form of tiotropium ameliorates some of the adverse effects of ICS treatment with the incidence rates returning towards those patients taking other ICS. We suggest that the use of ICS without LAMA may expose the patient to additional risk of respiratory adverse events.
We conducted the analyses on the classic basis of intention to treat despite this our observations have strengths and weaknesses. This large study population was observed over a 4-year period in patients with a range of airflow limitation and COPD severity. Co-existing treatment with ICS was uninfluenced either at study entry or during the course of the study. This avoids a major source of bias. A downside of this analysis is that we have not corrected for withdrawal or changing of ICS during the course of the study. In this regard, we are conducting the analysis on the basis of ‘intention to treat’. A further limitation of this study is the reliance on secondary endpoints which were less tightly defined. Thus, exacerbations of COPD were defined as an increase in new onset of at least or more symptoms for >3 days which required additional treatment with antibiotics and/or corticosteroids. It is likely therefore that some cases of pneumonia were included in this loose definition. Pneumonia was only defined as an investigator-reported adverse event not necessarily requiring confirmation by chest X-ray. In contrast to several large database studies with similar findings, the prospective nature and rigid diagnostic criteria for entry into UPLIFT ensures that our analysis has been performed on a clearly defined COPD population.
Compliance with Ethical Standards